The power of gene and variant-level expression in immuno-oncology biomarker discovery
Recent advancements in RNASeq technology have revolutionized the search for the genes and gene expression signatures that determine immunotherapy response by providing a high throughput method of comprehensively examining changes in these functional elements. RNASeq is now established as a cost effective, rapid approach to exploring a myriad of molecular insights including reliable detection of low abundant transcripts, gene fusion discovery, or expression of cancer-related gene pathways.
Transcriptomic features in both the tumor and tumor microenvironment
The significant value of using RNASeq and other RNA-based methodologies to better understand clinical response to checkpoint inhibitors is apparent across numerous recent studies. Let’s take a closer look at a few of these noteworthy publications:
- At the genomic level, Hugo et al., 2016 found that tumor mutational burden (TMB) is not predictive of patient response to anti-PD-1 treatment in melanoma patients. However, the authors subsequently investigated whether transcriptomic features provide clues which correlate with clinical response. Using RNASeq, differential expression analysis revealed an enrichment of signatures, termed as innate anti-PD-1 resistance or IPRES signature. The authors found mesenchymal genes as well as angiogenesis, hypoxia, and wound-healing genes were expressed at higher levels in non-responding patient tumor samples when compared to anti-PD-1 blockade-responsive patients.
- Tumor PD-L1 expression by IHC can provide some clinical insight for predicting patient response, but it does not reveal the intricacies of tumor and immune cell interaction. Beginning with a small melanoma cohort of 19 patient tumor biopsies, Ayers et al., 2017 refined a gene expression profile of a subset of genes associated with a T cell-inflamed microenvironment that predict response to PD-1 blockade. The authors created a targeted list of IFN-y signaling genes and an expanded immune signature gene set of cytokines, cell markers and immunomodulatory factors. Their analysis also extended to various other cancer types such as gastric cancer and head and neck squamous cell carcinoma. Further evaluations using this predictive gene signature are currently ongoing in pembrolizumab clinical trials.
- Published data from the group at Memorial Sloan Kettering Cancer Center (Riaz et al., 2017) showed interesting findings from RNASeq of both pre- and on-therapy biopsies in nivolumab-treated advanced melanoma patients. While numerous, differentially expressed genes were upregulated in on-treatment samples, more immune-related genes were upregulated specifically in responders. These included genes such as TNFRSF4 (OX40), HAVCR2 (TIM-3), and others involved in lymphocytic activation as being significantly over-expressed in patients responding to anti-PD-1 therapy.
RNASeq value in other IO applications
Gene-level expression changes are undeniably helping to shape our understanding of patient response to checkpoint inhibitors like nivolumab and pembrolizumab. Along with gene-level expression, variant-level expression data may also prove quite useful in immunotherapy development as well.
Currently, for personalized cancer vaccines, the most established criteria in the field is antigen-binding prediction when prioritizing putative neoantigens. For instance, algorithms provide a binding affinity regarding the MHC alleles that the predicted peptide may bind to in nanomolar concentration using NGS data from the tumor specimen. In general, less than a 500 nM binding affinity threshold is considered to be a decent binding strength for potential peptides to be recognized by T cells.
In one of the first-in-human studies using neoantigen vaccines published last year (Sahin et al., 2017), predicted HLA high-affinity binding as well as transcript and RNA variant allele frequency played a key role in the neo-epitope prioritization criteria used in selecting candidates. Thus, accuracy in both gene- and variant-level expression may provide strong predictive value in neoantigen-based cancer vaccine design.
A “power tool” in the toolbox
RNA data is a powerful tool for both immuno-oncology translational research and clinical trials. Determining whether or not a gene is mutated is no longer enough — especially in biomarker discovery for immunotherapy, due to the complex interplay between the tumor, its microenvironment, and immune response. While the studies summarized here provide only a glimpse into how RNASeq is revolutionizing immuno-oncology biomarker discovery, it is clear that having accurate gene and variant expression will meaningfully contribute to the ultimate goal of unveiling predictors of patient response.